Estimating the Green Line
Today we’ll explore my process of estimating my green line for a race, and deciding on the actual paces and splits to aim for during a race.
Photo credit Melissa Ruse
First, I’ll outline the process I use. Note that this process is iterative - I start following it in order but will jump back to previous steps as I have more time to digest the paces, and my fitness evolves with training.
As you’ll see, this is part science and part art - it relies on a lot of data, but also knowledge of my own abilities and potential. Honesty is critical here - I can have as much wishful thinking as I’d like but I will pay for over-aggressive estimations on race day.
Here are the high level steps.
Set a Baseline - I begin by getting a baseline of what I think I’m capable of at a race. This is the finish time of executing a perfect green line race at maximal ability. I use a few techniques and weigh them based on how accurate I believe they are.
Tweak - I holistically consider a wide array of other factors, tweaking the baseline up or down according to their influences.
Goal Setting - From the idealized finish time, I consider my goals and produce a desired race time.
Pace Plan - Based on my goals and desired finish time, I come up with a concrete set of paces or splits.
Contingencies - From the primary pace plan, I consider alternatives if the day goes better or worse than expected.
Visualization - I mentally prepare for all of the contingencies so I can fluidly adapt race day.
Repeat - Go through the above processes again, sanity checking and hardening confidence in the numbers.
Baseline - Previous Performances
The starting point for most of my paces are my previous performances at similar races or distances. There are few things more predictive of my future racing than my past racing. I will weigh each based on how recent they were, how similar my training and conditioning are to those races, and how much the terrain translates.
Example - Prepping for 100 miles at 6 days in the dome, I considered my 14:09 Javelina Jundred 8 months before, my 14:22 split at Desert Solstice 6 months before, and several 5:4X-5:5X 50 mile races. These plus my increased fitness indicated my ideal pace should be in the high 13:XX range.
Baseline - Percentage of World Records or Course Records
Next, I narrow in on a specific number using a percentage of world record chart. This works especially well for flat races. The same technique can be utilized with a course record, though course record times are usually not as hardened as world records so I generally estimate what an equivalent course record would be with world record competition.
Here is a sample calculator that I use, which can be copied to plug in your own times. For myself, I can see where each of my own PRs falls relative to the world record. I then eyeball and estimate what my potential is at the goal distance. I take into account how good or bad each PR is relative to my current fitness.
Example - For the Dome, I had PRs between the marathon and 24 hours ranging from 80-85%. I had similar fitness to my best PRs, so I looked at the 85% mark for 100 miles which was 13:14. This confirmed my initial baseline was in the ballpark, and provided an even more accurate target.
Callout - Not all world records are created equal. Records up to the marathon have been hardened by a lot of competition, money, and ability to optimize distances where fewer things go wrong. The world records get softer as they get longer in general, so there may be an argument for hitting a higher overall percentage of certain ultra distances. I’ve found in practice that this effect is counterbalanced by the difficulty in getting everything right in a race as distances increase. The record itself reflects the amount of things that typically go wrong, so is still accurate to rely on. My American Record at 24 hours was 4% above my closest PR largely due to this effect, but I nailed virtually every detail of execution to make it possible.
Second Callout - This method becomes less accurate as the percentage gets further from 100% when mixing fixed distance and fixed time predictions. This is because the percentage approach increases time on feet in one direction, but reduces distance in the other direction. For example, if the world record for 100 miles was 12 hours (and vice versa), then 50% predicts a 100 mile time of 24 hours, while 50% of 12 hours predicts 50 miles which is the same recommended pace for half the distance/time. It’s best to use just fixed distances or just fixed times if this becomes a significant factor.
Baseline - Trail Performances
Percent of world record often doesn’t translate well to trail races. Another technique I use is to look at historical results at other runners I know to be similar to my current abilities. This takes some research to understand if the course itself got easier or harder over the years, what kind of shape each runner was in any given year, and what kind of day they had that race. But after looking at a dozen data points, I get a general sense of what a realistic time should be for me.
Example - Run Rabbit Run 100 this year, I looked at performances from Jason Schlarb,, Kyle Pietari, Alex Nichols, Mark Hammond, Geoff Browning, and others who were all near my ability. Some of the years were on a faster course, and some were on a slightly slower course. Looking at all the performances which ranged from 17:XX for Schlarb’s best year to 19:XX for some of the others, I estimated my fitness could put me in the high 18:XX range.
Tweak - Current Fitness
To a certain extent, the baselines above inherently take into account my current fitness - I’m already considering it at each step. But I take extra time to consider the nuances of my fitness today. If I visualize the paces needed to hit the finish time, especially the last 40% of the race, does it sound doable or does it sound challenging knowing how I feel during races?
Are my percent of world record numbers in this training cycle higher than any historical PR? Am I on an upward trajectory and continuing to improve steadily? Or am I plateauing without a good indicator I’m ready to push higher? Does my current fitness indicate I’m getting stronger as races get longer and require more endurance, or does my fitness only really shine at shorter distances?
I try to get objective about this as much as possible, and distrust my memory of my previous training cycles unless supported by data. I look at comparable races year over year, and weekly mileage trends between years. I’ll often also repeat key workouts and training runs to see how my speed and endurance are trending. For years I have done 50 mile flat training runs or races as a fitness checkpoint, and it proves quite predictive of my year over year improvements (or regressions).
Tweak - Past Races
I consider my own past planning of races to learn meta-lessons about my ability to predict. When I previously thought a finish time was realistic, did I blow up and slow down, or finish strong? I want a high degree of confidence the race paces are achievable, and not just easy sounding on paper. What evidence do I have that my current level of fitness translates to racing success based on past fitness at past races?
Tweak - Trail from Flat
If I’m estimating a trail race, I can still come up with a flat estimate for the same amount of time on feet. I can then difficulty adjust the flat pace to the trail pace to see if the effort sounds correct.
Example - At Run Rabbit Run I expected to finish in around 19 hours. My pace estimate for a 19 hour flat race would be 8:30 per mile. The RRR pace for ~104 trail miles is ~11:00 per mile. This doesn’t mean a lot yet, but when setting paces on specific segments I will sanity check each climb and descent against the 8:30 effort and get an idea if that is realistic or too aggressive for trail - and adjust the finish time accordingly. Is a 13 minute mile realistic for a 500 foot/mile climb at elevation? I can simulate it in training and see if it is a similar effort to 8:30/mile on flat.
Tweak - Experience Increase
So far, I have primarily relied on backward-looking data points and historical data, and have avoided setting a higher goal based on new potential. This is for a very specific reason. It’s easy to set big goals with wishful thinking, but most of us know we can easily fall short of those goals. Think back on your own wishful goals for ultras, and consider how frequently you’ve hit them vs missed them.
I now take a moment to consider if I’m on an upward trajectory of hitting my goals. If I am hitting or exceeding my planned finish times at 70-90% of my races, I gain overall confidence in my ability to estimate. This will let me be a bit more aggressive (1-3%) in how close my actual goal will be to the maximal baseline. If I’m missing goals at 30% or more of my races, I should expect that setting higher goals will substantially increase the chance I’ll miss them.
It’s worth mentioning that the benefit of green line racing is that we’re unlikely to miss out on a bigger goal being realistic - it just means we’ll hit a bigger finishing kick, make up the time later in the race, and then be able to set a higher goal pace next time.
Example - With practice, I increased how often I hit my goal times and negative splits from 2017 to present. In 2021, I was hitting my goals 80-90% of the time. This gave me confidence to push a bit higher than projected at Desert Solstice as I had built up “muscle memory” of precisely pacing early on and increasing the pace later on, especially under fatigue.
Tweak - Other Factors
Finally, I have a large bucket of other factors to consider. This can range from improved nutrition to better logistics to sports psychology improvements. It could be training with and adopting super shoes, fewer and faster stops at aid stations, or better heat mitigation.
I don’t leave the effects of these to chance. For the most part I practice them in races or time trials in the buildup, and they automatically get factored in through the baselining process. But I also take time to consider anything that may not have been captured through this process.
Example - At Desert Solstice in 2021, all of my baselining races were done either on fatigued legs or as training races. Most of them were self-crewed. I had been steadily getting better at the psychology of long ultras and would also be breaking out a new pair of super shoes unlike the buildup races. All this together let me bump up my baseline just a bit.
Tweak - Weaknesses
The counterpoint to experience and other factors are known weaknesses, and these are just as important to consider. Perhaps I have no problem executing through 100k or 100 miles, but I know I will struggle past 18 hours in a 24 hour run. The baselining process may not capture this, and I should be willing to adjust my baseline down in response to make it realistic.
There will still be options to overcome and exceed this later on - but for now I want to maximize the realistic time target.
Setting the Pace Plan
Now that we have a baseline of what we believe is realistically possible, it’s time to get down to the details of creating a pace plan for the race.
Deciding on Goals
The first step to getting a pace plan is to decide the goal finish time based on the baseline. This will be based on risk tolerance and desired recovery. See Part 1 for discussion of these tradeoffs.
For most of my races, my goal pace is 5-10% below my baseline. This gives me a healthy margin to negative split while recovering quickly, but also keeps the door open to negative splitting significantly on a good day and getting a lot closer to the baseline.
Example - My baseline at the Dome was 13:00-13:10. My race plan was 13:45-13:55 which is about 6% slower. I ran a 3% negative split to run 13:20 after the day went well.
Ah, spreadsheets! Everyone’s favorite reason to get out and run!
This may not be the fun part, but it’s one of the places I spend the most time tweaking using all of the above principles. I’ll provide some examples of my own spreadsheets used for different races, which vary based on what factors I think are most important to estimating paces for that race. Please copy them if they are useful as starting points.
I start with even splits (difficulty adjusted) for my goal time. I then tweak bit by bit based on the ideas in the Part 1 and Part 2 blog posts (especially the green line), plus everything else here. I re-ask all the questions I’ve posed while looking at the splits and visualizing as I’ll describe later.
As a reminder, I calculate negative/positive splits with one of these formulas:
split = (second half minutes - first half minutes) / total minutes
split = (first half distance - second half distance) / total distance
split = (second half pace - first half pace) / (total pace * 2)
Tweak - Trail
What about on trails where the numbers aren’t obvious? I have a few techniques here too. My first pass, I look for a pivot point near the middle of the course that splits the race in two with roughly equal difficulty per mile. Is there an aid station near halfway at similar elevation to the start/finish, and with similar difficulty climbs on the two halves? This lets me estimate the negative split based on the average pace without needing two perfectly even halves. This will create a general shape for the race strategy.
Next, I throw in extremely rough cut times for each segment. I then go deeper on a process described earlier - analyzing athletes similar to me. I take the group that I used to baseline and now start looking at both their individual segment splits as well as their positive/negative splits based on the pivot point. I get data from anywhere I can. If the race doesn’t provide historical splits, I’ve scraped Strava activities by finding the exact times they arrived at aid stations. Once I have good data, I take into account the same factors as before - their fitness that year, how strong or weak they were in the second half, and any other relevant information. I continually adjust my splits based on the comparison data and my own strengths and weaknesses on uphills, downhills, and technical stretches. I expect my first half splits to be slower than the comparison splits, and my second half splits to be increasingly faster. If the runner had a yellow line race, I still expect to be faster at the very end but by less than 60-90% into the race.
Once I have my goal splits, I chart out contingency splits. What if I have a stellar day? What if I have a terrible day? I create upper and lower bound strategies that usually look exactly the same for the first half, but then slowly drift faster or slower from halfway to the finish. I often use a 3-5% negative split for the upper bound and a 1-5% positive split for a lower bound.
With the pace chart created, I now spend time visualizing the race using the paces. I walk through the first half in my head, keeping everything calm and under control but also crisp and forward-pushing. I focus on flowing at the pace while fully conserving and preserving myself.
I visualize the second half of the race, imagining the fatigue I will feel at each point, and how I will still have plenty left to hold my pace. I imagine slowly letting myself get more excited as I get closer and closer to the finish, inching up the pace, passing more and more runners and enjoying the crescendo I’m creating.
I’ll run through the second half and how I’ll feel on a better than expected day, pushing faster than planned at the very end and imagining the joy I’ll feel. I go through the emotions of a bad day, where things don’t go well, I fall off pace but still keep my attitude positive. I imagine still enjoying the privilege of being a part of the race, still getting in a good run, and making it to the finish all the same.
Alongside the mental visualizations, I’m utilizing physical visualizations. I do a variety of runs to simulate different parts of the race. I tackle similar climbs at the planned race pace to know how they will feel. I create fatigue scenarios to simulate later parts of the race. At each point taking my mental visualizations and putting myself into the same head space I expect to be during the race on actual runs.
As mentioned multiple times, I cycle through some or all parts of this process over and over again until I have confidence that the plan is where I want it and my confidence in executing it on race day is high. The more important the race, the more cycles I will run through and the more weeks or months I will spend on the process.
Most of what I’ve described is aimed to be extremely realistic and avoid shooting for aspirational goals. But there is an important way to utilize this to reach goals that are known to be just a bit beyond reach. By starting this process far enough before the goal race, it will highlight how unrealistic the goal is and where the gaps are. It then becomes a roadmap of the improvements to fitness and other factors that are needed to make it realistic. Can I train smarter and harder to increase my predictive performances along the way and indicate my fitness is increasing? Can I shorten aid station times to make faster splits more achievable? Can I identify the factors that make me fade late in a race and improve them? Can I increase my confidence in late race speed by practicing in training races?
Example - This is the technique I used in 2021 preparing for my 24 hour American Record at Desert Solstice. When I started this process in January, my estimates put me at 165-168 miles. I worked incrementally, inching closer and closer every few months until race day, when my estimations told me I was in the 171-173 range.
As with my other posts, I would love to hear how helpful this is. What other experiences can I share that would be helpful? What parts of negative splitting aren’t obvious or easy? Leave a comment or send me a message, if there are a lot of good ones I’d like to do a dedicated post answering any questions!